Operative Time Prediction by Machine Learning for Robot-Assisted Laparoscopic Radical Prostatectomy

Area of Science:

• Urology
• Surgical Oncology
• Data Science

Background:

• Efficient operating room scheduling is vital for healthcare systems.
• Accurate prediction of operative time for robot-assisted radical prostatectomy (RARP) remains a challenge.
• Optimizing surgical workflow requires precise time estimations.

Purpose of the Study:

• To develop and validate an operative time prediction system for RARP.
• To enhance the accuracy of predicting surgical duration in robotic prostatectomy.
• To provide a tool for improving operating room scheduling efficiency.

Main Methods:

• Retrospective analysis of 557 (Tohoku University Hospital) and 150 (Miyagi Cancer Center) RARP patients.
• Collected variables included patient demographics, comorbidities, tumor characteristics, and surgical factors.
• Developed an integrated prediction system using approximation curves and a random forest machine learning model.

Main Results:

• Achieved a normalized root mean square error of 0.107 (internal) and 0.148 (external) validation.
• Demonstrated superior reliability compared to operator-based time estimations.
• Identified key predictive factors: lymphadenectomy, grade group, prostate volume, and body mass index.

Conclusions:

• The developed system offers reliable and robust operative time predictions for RARP.
• The system effectively incorporates known factors influencing surgical duration.
• Potential to significantly improve operating room scheduling and resource management.